From 662a59d0fda3f763a2f67654dd50cc2cf38a8178 Mon Sep 17 00:00:00 2001 From: Chad Fraleigh Date: Mon, 14 Aug 2023 19:21:30 -0700 Subject: [PATCH 1/7] Eliminate additional undefined behavior of bit shifting signed int. --- libi2pd/Base.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libi2pd/Base.cpp b/libi2pd/Base.cpp index 0e69b663..e979c5b4 100644 --- a/libi2pd/Base.cpp +++ b/libi2pd/Base.cpp @@ -272,7 +272,7 @@ namespace data size_t Base32ToByteStream (const char * inBuf, size_t len, uint8_t * outBuf, size_t outLen) { - int tmp = 0, bits = 0; + unsigned int tmp = 0, bits = 0; size_t ret = 0; for (size_t i = 0; i < len; i++) { From c112276eea649c3b406953e3400573d78a049f20 Mon Sep 17 00:00:00 2001 From: orignal Date: Sat, 26 Aug 2023 10:57:05 -0400 Subject: [PATCH 2/7] memory poll for RouterProfile --- libi2pd/NetDb.cpp | 1 + libi2pd/NetDb.hpp | 2 ++ libi2pd/Profiling.cpp | 3 ++- libi2pd/Profiling.h | 3 +-- 4 files changed, 6 insertions(+), 3 deletions(-) diff --git a/libi2pd/NetDb.cpp b/libi2pd/NetDb.cpp index 061ccc23..02d4dfd8 100644 --- a/libi2pd/NetDb.cpp +++ b/libi2pd/NetDb.cpp @@ -175,6 +175,7 @@ namespace data if (ts - lastProfilesCleanup >= (uint64_t)(i2p::data::PEER_PROFILE_AUTOCLEAN_TIMEOUT + profilesCleanupVariance) || ts + i2p::data::PEER_PROFILE_AUTOCLEAN_TIMEOUT < lastProfilesCleanup) { + m_RouterProfilesPool.CleanUpMt (); if (m_PersistProfiles) PersistProfiles (); DeleteObsoleteProfiles (); lastProfilesCleanup = ts; diff --git a/libi2pd/NetDb.hpp b/libi2pd/NetDb.hpp index b7d9a5b3..44dff3b8 100644 --- a/libi2pd/NetDb.hpp +++ b/libi2pd/NetDb.hpp @@ -128,6 +128,7 @@ namespace data }; std::shared_ptr NewLease (const Lease& lease) { return m_LeasesPool.AcquireSharedMt (lease); }; std::shared_ptr NewIdentity (const uint8_t * buf, size_t len) { return m_IdentitiesPool.AcquireSharedMt (buf, len); }; + std::shared_ptr NewRouterProfile () { return m_RouterProfilesPool.AcquireSharedMt (); }; uint32_t GetPublishReplyToken () const { return m_PublishReplyToken; }; @@ -185,6 +186,7 @@ namespace data i2p::util::MemoryPoolMt m_RouterInfoAddressVectorsPool; i2p::util::MemoryPoolMt m_LeasesPool; i2p::util::MemoryPoolMt m_IdentitiesPool; + i2p::util::MemoryPoolMt m_RouterProfilesPool; }; extern NetDb netdb; diff --git a/libi2pd/Profiling.cpp b/libi2pd/Profiling.cpp index 879aea29..2031fa39 100644 --- a/libi2pd/Profiling.cpp +++ b/libi2pd/Profiling.cpp @@ -16,6 +16,7 @@ #include "FS.h" #include "Log.h" #include "Timestamp.h" +#include "NetDb.hpp" #include "Profiling.h" namespace i2p @@ -237,7 +238,7 @@ namespace data if (it != g_Profiles.end ()) return it->second; } - auto profile = std::make_shared (); + auto profile = netdb.NewRouterProfile (); profile->Load (identHash); // if possible std::unique_lock l(g_ProfilesMutex); g_Profiles.emplace (identHash, profile); diff --git a/libi2pd/Profiling.h b/libi2pd/Profiling.h index c351b41d..6531e060 100644 --- a/libi2pd/Profiling.h +++ b/libi2pd/Profiling.h @@ -31,7 +31,7 @@ namespace data const char PEER_PROFILE_USAGE_CONNECTED[] = "connected"; const int PEER_PROFILE_EXPIRATION_TIMEOUT = 36; // in hours (1.5 days) - const int PEER_PROFILE_AUTOCLEAN_TIMEOUT = 6 * 3600; // in seconds (6 hours) + const int PEER_PROFILE_AUTOCLEAN_TIMEOUT = 3 * 3600; // in seconds (3 hours) const int PEER_PROFILE_AUTOCLEAN_VARIANCE = 3600; // in seconds (1 hour) const int PEER_PROFILE_DECLINED_RECENTLY_INTERVAL = 150; // in seconds (2.5 minutes) const int PEER_PROFILE_PERSIST_INTERVAL = 3300; // in seconds (55 minutes) @@ -43,7 +43,6 @@ namespace data public: RouterProfile (); - RouterProfile& operator= (const RouterProfile& ) = default; void Save (const IdentHash& identHash); void Load (const IdentHash& identHash); From 32c5ff23a62b8b8f200b4717244524bb3ab68638 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E0=B0=9C=E0=B0=BF=E0=B0=82=E0=B0=A6=E0=B0=82=20=E0=B0=B5?= =?UTF-8?q?=E0=B0=BE=E0=B0=90=E0=B0=BF?= <109575325+jindam-vani@users.noreply.github.com> Date: Thu, 31 Aug 2023 16:51:21 +0000 Subject: [PATCH 3/7] Differentiate cryptocurrency & its associated address (#1951) --- README.md | 24 +++++++++++++++++------- 1 file changed, 17 insertions(+), 7 deletions(-) diff --git a/README.md b/README.md index d0481ed9..e7fb7318 100644 --- a/README.md +++ b/README.md @@ -99,13 +99,23 @@ Current status: [![Crowdin](https://badges.crowdin.net/i2pd/localized.svg)](http Donations --------- -BTC: 3MDoGJW9TLMTCDGrR9bLgWXfm6sjmgy86f -LTC: LKQirrYrDeTuAPnpYq5y7LVKtywfkkHi59 -ETH: 0x9e5bac70d20d1079ceaa111127f4fb3bccce379d -DASH: Xw8YUrQpYzP9tZBmbjqxS3M97Q7v3vJKUF -ZEC: t1cTckLuXsr1dwVrK4NDzfhehss4NvMadAJ -GST: GbD2JSQHBHCKLa9WTHmigJRpyFgmBj4woG -XMR: 497pJc7X4xqKvcLBLpSUtRgWqMMyo24u4btCos3cak6gbMkpobgSU6492ztUcUBghyeHpYeczB55s38NpuHoH5WGNSPDRMH +**E-Mail**: ```i2porignal at yandex.ru``` + +**BTC**: ```3MDoGJW9TLMTCDGrR9bLgWXfm6sjmgy86f``` + +**LTC**: ```LKQirrYrDeTuAPnpYq5y7LVKtywfkkHi59``` + +**ETH**: ```0x9e5bac70d20d1079ceaa111127f4fb3bccce379d``` + +**GST**: ```GbD2JSQHBHCKLa9WTHmigJRpyFgmBj4woG``` + +**DASH**: ```Xw8YUrQpYzP9tZBmbjqxS3M97Q7v3vJKUF``` + +**ZEC**: ```t1cTckLuXsr1dwVrK4NDzfhehss4NvMadAJ``` + +**ANC**: ```AQJYweYYUqM1nVfLqfoSMpUMfzxvS4Xd7z``` + +**XMR**: ```497pJc7X4xqKvcLBLpSUtRgWqMMyo24u4btCos3cak6gbMkpobgSU6492ztUcUBghyeHpYeczB55s38NpuHoH5WGNSPDRMH``` License ------- From 7b6aa41ca8dc3ca2a0c9d01ff9870794ad98e403 Mon Sep 17 00:00:00 2001 From: R4SAS Date: Thu, 31 Aug 2023 16:52:51 +0000 Subject: [PATCH 4/7] CPU: remove AVX code, switch to __builtin for AES detection (#1959) * [cpu] remove avx detect and code blocks, try to switch to __builtin * [cpu] use __builtin_* only on x86 systems * [cpu] perform check in separate function * [cpu] set AES definition on MSVC * update x86 and aes support checks at compile time * [cmake] update comment about AES on MSVC --- .editorconfig | 3 + build/CMakeLists.txt | 11 +- contrib/i2pd.conf | 2 - daemon/Daemon.cpp | 3 +- libi2pd/CPU.cpp | 54 +++---- libi2pd/CPU.h | 5 +- libi2pd/Config.cpp | 4 +- libi2pd/Crypto.cpp | 332 ++++++++++++++++++++++--------------------- libi2pd/Crypto.h | 2 +- libi2pd/Identity.cpp | 29 +--- libi2pd/api.cpp | 3 +- 11 files changed, 219 insertions(+), 229 deletions(-) diff --git a/.editorconfig b/.editorconfig index e1f1243a..0f57a84a 100644 --- a/.editorconfig +++ b/.editorconfig @@ -34,3 +34,6 @@ trim_trailing_whitespace = false [*.yml] indent_style = space indent_size = 2 + +[*.patch] +trim_trailing_whitespace = false diff --git a/build/CMakeLists.txt b/build/CMakeLists.txt index f5d01a0b..3185ffab 100644 --- a/build/CMakeLists.txt +++ b/build/CMakeLists.txt @@ -197,14 +197,11 @@ endif() # Note: AES-NI and AVX is available on x86-based CPU's. # Here also ARM64 implementation, but currently we don't support it. -# MSVC is not supported. -if(MSVC) - message(STATUS "AES-NI is not supported on MSVC, option was disabled") - set(WITH_AESNI OFF) -endif() - +# MSVC is not supported due to different ASM processing, so we hope OpenSSL has its own checks to run optimized code. if(WITH_AESNI AND (ARCHITECTURE MATCHES "x86_64" OR ARCHITECTURE MATCHES "i386")) - set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -maes") + if(NOT MSVC) + set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -maes") + endif() add_definitions(-D__AES__) endif() diff --git a/contrib/i2pd.conf b/contrib/i2pd.conf index 5facbbd5..545973a3 100644 --- a/contrib/i2pd.conf +++ b/contrib/i2pd.conf @@ -280,8 +280,6 @@ verify = true [cpuext] ## Use CPU AES-NI instructions set when work with cryptography when available (default: true) # aesni = true -## Use CPU AVX instructions set when work with cryptography when available (default: true) -# avx = true ## Force usage of CPU instructions set, even if they not found (default: false) ## DO NOT TOUCH that option if you really don't know what are you doing! # force = false diff --git a/daemon/Daemon.cpp b/daemon/Daemon.cpp index 8e9721db..f04236fe 100644 --- a/daemon/Daemon.cpp +++ b/daemon/Daemon.cpp @@ -150,12 +150,11 @@ namespace util bool precomputation; i2p::config::GetOption("precomputation.elgamal", precomputation); bool aesni; i2p::config::GetOption("cpuext.aesni", aesni); - bool avx; i2p::config::GetOption("cpuext.avx", avx); bool forceCpuExt; i2p::config::GetOption("cpuext.force", forceCpuExt); bool ssu; i2p::config::GetOption("ssu", ssu); if (!ssu && i2p::config::IsDefault ("precomputation.elgamal")) precomputation = false; // we don't elgamal table if no ssu, unless it's specified explicitly - i2p::crypto::InitCrypto (precomputation, aesni, avx, forceCpuExt); + i2p::crypto::InitCrypto (precomputation, aesni, forceCpuExt); i2p::transport::InitAddressFromIface (); // get address4/6 from interfaces diff --git a/libi2pd/CPU.cpp b/libi2pd/CPU.cpp index 0804e2ac..d42f1ddd 100644 --- a/libi2pd/CPU.cpp +++ b/libi2pd/CPU.cpp @@ -1,5 +1,5 @@ /* -* Copyright (c) 2013-2020, The PurpleI2P Project +* Copyright (c) 2013-2023, The PurpleI2P Project * * This file is part of Purple i2pd project and licensed under BSD3 * @@ -7,52 +7,52 @@ */ #include "CPU.h" -#if defined(__x86_64__) || defined(__i386__) -#include -#endif #include "Log.h" +#if defined(_MSC_VER) +#include + #ifndef bit_AES -#define bit_AES (1 << 25) + #define bit_AES (1 << 25) #endif -#ifndef bit_AVX -#define bit_AVX (1 << 28) #endif - namespace i2p { namespace cpu { bool aesni = false; - bool avx = false; - void Detect(bool AesSwitch, bool AvxSwitch, bool force) + inline bool cpu_support_aes() { -#if defined(__x86_64__) || defined(__i386__) - int info[4]; - __cpuid(0, info[0], info[1], info[2], info[3]); - if (info[0] >= 0x00000001) { - __cpuid(0x00000001, info[0], info[1], info[2], info[3]); -#if defined (_WIN32) && (WINVER == 0x0501) // WinXP - if (AesSwitch && force) { // only if forced +#if (defined(_M_AMD64) || defined(__x86_64__)) || (defined(_M_IX86) || defined(__i386__)) +#if defined(_MSC_VER) + int cpu_info[4]; + __cpuid(cpu_info, 1); + return ((cpu_info[2] & bit_AES) != 0); +#elif defined(__clang__) +#if __clang_major__ >= 6 + __builtin_cpu_init(); +#endif + return __builtin_cpu_supports("aes"); +#elif defined(__GNUC__) + __builtin_cpu_init(); + return __builtin_cpu_supports("aes"); #else - if ((info[2] & bit_AES && AesSwitch) || (AesSwitch && force)) { + return false; #endif - aesni = true; - } -#if defined (_WIN32) && (WINVER == 0x0501) // WinXP - if (AvxSwitch && force) { // only if forced #else - if ((info[2] & bit_AVX && AvxSwitch) || (AvxSwitch && force)) { + return false; #endif - avx = true; - } + } + + void Detect(bool AesSwitch, bool force) + { + if ((cpu_support_aes() && AesSwitch) || (AesSwitch && force)) { + aesni = true; } -#endif // defined(__x86_64__) || defined(__i386__) LogPrint(eLogInfo, "AESNI ", (aesni ? "enabled" : "disabled")); - LogPrint(eLogInfo, "AVX ", (avx ? "enabled" : "disabled")); } } } diff --git a/libi2pd/CPU.h b/libi2pd/CPU.h index f021bccb..a0695884 100644 --- a/libi2pd/CPU.h +++ b/libi2pd/CPU.h @@ -1,5 +1,5 @@ /* -* Copyright (c) 2013-2020, The PurpleI2P Project +* Copyright (c) 2013-2023, The PurpleI2P Project * * This file is part of Purple i2pd project and licensed under BSD3 * @@ -14,9 +14,8 @@ namespace i2p namespace cpu { extern bool aesni; - extern bool avx; - void Detect(bool AesSwitch, bool AvxSwitch, bool force); + void Detect(bool AesSwitch, bool force); } } diff --git a/libi2pd/Config.cpp b/libi2pd/Config.cpp index 6b515ef9..5551b010 100644 --- a/libi2pd/Config.cpp +++ b/libi2pd/Config.cpp @@ -193,7 +193,7 @@ namespace config { options_description precomputation("Precomputation options"); precomputation.add_options() ("precomputation.elgamal", -#if defined(__x86_64__) +#if (defined(_M_AMD64) || defined(__x86_64__)) value()->default_value(false), #else value()->default_value(true), @@ -308,7 +308,7 @@ namespace config { options_description cpuext("CPU encryption extensions options"); cpuext.add_options() ("cpuext.aesni", bool_switch()->default_value(true), "Use auto detection for AESNI CPU extensions. If false, AESNI will be not used") - ("cpuext.avx", bool_switch()->default_value(true), "Use auto detection for AVX CPU extensions. If false, AVX will be not used") + ("cpuext.avx", bool_switch()->default_value(false), "Deprecated option") ("cpuext.force", bool_switch()->default_value(false), "Force usage of CPU extensions. Useful when cpuinfo is not available on virtual machines") ; diff --git a/libi2pd/Crypto.cpp b/libi2pd/Crypto.cpp index 12087443..b5bc33d1 100644 --- a/libi2pd/Crypto.cpp +++ b/libi2pd/Crypto.cpp @@ -1,5 +1,5 @@ /* -* Copyright (c) 2013-2022, The PurpleI2P Project +* Copyright (c) 2013-2023, The PurpleI2P Project * * This file is part of Purple i2pd project and licensed under BSD3 * @@ -28,6 +28,12 @@ #include "I2PEndian.h" #include "Log.h" +#if defined(__AES__) && !defined(_MSC_VER) && ((defined(_M_AMD64) || defined(__x86_64__)) || (defined(_M_IX86) || defined(__i386__))) + #define SUPPORTS_AES 1 +#else + #define SUPPORTS_AES 0 +#endif + namespace i2p { namespace crypto @@ -361,7 +367,7 @@ namespace crypto BIGNUM * b1 = BN_CTX_get (ctx); BIGNUM * b = BN_CTX_get (ctx); // select random k -#if defined(__x86_64__) +#if (defined(_M_AMD64) || defined(__x86_64__)) BN_rand (k, ELGAMAL_FULL_EXPONENT_NUM_BITS, -1, 1); // full exponent for x64 #else BN_rand (k, ELGAMAL_SHORT_EXPONENT_NUM_BITS, -1, 1); // short exponent of 226 bits @@ -428,7 +434,7 @@ namespace crypto void GenerateElGamalKeyPair (uint8_t * priv, uint8_t * pub) { -#if defined(__x86_64__) || defined(__i386__) || defined(_MSC_VER) +#if (defined(_M_AMD64) || defined(__x86_64__)) || (defined(_M_IX86) || defined(__i386__)) || defined(_MSC_VER) RAND_bytes (priv, 256); #else // lower 226 bits (28 bytes and 2 bits) only. short exponent @@ -555,7 +561,7 @@ namespace crypto } // AES -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES #define KeyExpansion256(round0,round1) \ "pshufd $0xff, %%xmm2, %%xmm2 \n" \ "movaps %%xmm1, %%xmm4 \n" \ @@ -580,7 +586,7 @@ namespace crypto "movaps %%xmm3, "#round1"(%[sched]) \n" #endif -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES void ECBCryptoAESNI::ExpandKey (const AESKey& key) { __asm__ @@ -621,7 +627,7 @@ namespace crypto #endif -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES #define EncryptAES256(sched) \ "pxor (%["#sched"]), %%xmm0 \n" \ "aesenc 16(%["#sched"]), %%xmm0 \n" \ @@ -642,16 +648,18 @@ namespace crypto void ECBEncryption::Encrypt (const ChipherBlock * in, ChipherBlock * out) { -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES if(i2p::cpu::aesni) { __asm__ - ( - "movups (%[in]), %%xmm0 \n" - EncryptAES256(sched) - "movups %%xmm0, (%[out]) \n" - : : [sched]"r"(GetKeySchedule ()), [in]"r"(in), [out]"r"(out) : "%xmm0", "memory" - ); + ( + "movups (%[in]), %%xmm0 \n" + EncryptAES256(sched) + "movups %%xmm0, (%[out]) \n" + : + : [sched]"r"(GetKeySchedule ()), [in]"r"(in), [out]"r"(out) + : "%xmm0", "memory" + ); } else #endif @@ -660,7 +668,7 @@ namespace crypto } } -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES #define DecryptAES256(sched) \ "pxor 224(%["#sched"]), %%xmm0 \n" \ "aesdec 208(%["#sched"]), %%xmm0 \n" \ @@ -681,16 +689,18 @@ namespace crypto void ECBDecryption::Decrypt (const ChipherBlock * in, ChipherBlock * out) { -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES if(i2p::cpu::aesni) { __asm__ - ( - "movups (%[in]), %%xmm0 \n" - DecryptAES256(sched) - "movups %%xmm0, (%[out]) \n" - : : [sched]"r"(GetKeySchedule ()), [in]"r"(in), [out]"r"(out) : "%xmm0", "memory" - ); + ( + "movups (%[in]), %%xmm0 \n" + DecryptAES256(sched) + "movups %%xmm0, (%[out]) \n" + : + : [sched]"r"(GetKeySchedule ()), [in]"r"(in), [out]"r"(out) + : "%xmm0", "memory" + ); } else #endif @@ -699,7 +709,7 @@ namespace crypto } } -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES #define CallAESIMC(offset) \ "movaps "#offset"(%[shed]), %%xmm0 \n" \ "aesimc %%xmm0, %%xmm0 \n" \ @@ -708,7 +718,7 @@ namespace crypto void ECBEncryption::SetKey (const AESKey& key) { -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES if(i2p::cpu::aesni) { ExpandKey (key); @@ -722,28 +732,30 @@ namespace crypto void ECBDecryption::SetKey (const AESKey& key) { -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES if(i2p::cpu::aesni) { ExpandKey (key); // expand encryption key first // then invert it using aesimc __asm__ - ( - CallAESIMC(16) - CallAESIMC(32) - CallAESIMC(48) - CallAESIMC(64) - CallAESIMC(80) - CallAESIMC(96) - CallAESIMC(112) - CallAESIMC(128) - CallAESIMC(144) - CallAESIMC(160) - CallAESIMC(176) - CallAESIMC(192) - CallAESIMC(208) - : : [shed]"r"(GetKeySchedule ()) : "%xmm0", "memory" - ); + ( + CallAESIMC(16) + CallAESIMC(32) + CallAESIMC(48) + CallAESIMC(64) + CallAESIMC(80) + CallAESIMC(96) + CallAESIMC(112) + CallAESIMC(128) + CallAESIMC(144) + CallAESIMC(160) + CallAESIMC(176) + CallAESIMC(192) + CallAESIMC(208) + : + : [shed]"r"(GetKeySchedule ()) + : "%xmm0", "memory" + ); } else #endif @@ -754,28 +766,28 @@ namespace crypto void CBCEncryption::Encrypt (int numBlocks, const ChipherBlock * in, ChipherBlock * out) { -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES if(i2p::cpu::aesni) { __asm__ - ( - "movups (%[iv]), %%xmm1 \n" - "1: \n" - "movups (%[in]), %%xmm0 \n" - "pxor %%xmm1, %%xmm0 \n" - EncryptAES256(sched) - "movaps %%xmm0, %%xmm1 \n" - "movups %%xmm0, (%[out]) \n" - "add $16, %[in] \n" - "add $16, %[out] \n" - "dec %[num] \n" - "jnz 1b \n" - "movups %%xmm1, (%[iv]) \n" - : - : [iv]"r"((uint8_t *)m_LastBlock), [sched]"r"(m_ECBEncryption.GetKeySchedule ()), - [in]"r"(in), [out]"r"(out), [num]"r"(numBlocks) - : "%xmm0", "%xmm1", "cc", "memory" - ); + ( + "movups (%[iv]), %%xmm1 \n" + "1: \n" + "movups (%[in]), %%xmm0 \n" + "pxor %%xmm1, %%xmm0 \n" + EncryptAES256(sched) + "movaps %%xmm0, %%xmm1 \n" + "movups %%xmm0, (%[out]) \n" + "add $16, %[in] \n" + "add $16, %[out] \n" + "dec %[num] \n" + "jnz 1b \n" + "movups %%xmm1, (%[iv]) \n" + : + : [iv]"r"((uint8_t *)m_LastBlock), [sched]"r"(m_ECBEncryption.GetKeySchedule ()), + [in]"r"(in), [out]"r"(out), [num]"r"(numBlocks) + : "%xmm0", "%xmm1", "cc", "memory" + ); } else #endif @@ -799,22 +811,22 @@ namespace crypto void CBCEncryption::Encrypt (const uint8_t * in, uint8_t * out) { -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES if(i2p::cpu::aesni) { __asm__ - ( - "movups (%[iv]), %%xmm1 \n" - "movups (%[in]), %%xmm0 \n" - "pxor %%xmm1, %%xmm0 \n" - EncryptAES256(sched) - "movups %%xmm0, (%[out]) \n" - "movups %%xmm0, (%[iv]) \n" - : - : [iv]"r"((uint8_t *)m_LastBlock), [sched]"r"(m_ECBEncryption.GetKeySchedule ()), - [in]"r"(in), [out]"r"(out) - : "%xmm0", "%xmm1", "memory" - ); + ( + "movups (%[iv]), %%xmm1 \n" + "movups (%[in]), %%xmm0 \n" + "pxor %%xmm1, %%xmm0 \n" + EncryptAES256(sched) + "movups %%xmm0, (%[out]) \n" + "movups %%xmm0, (%[iv]) \n" + : + : [iv]"r"((uint8_t *)m_LastBlock), [sched]"r"(m_ECBEncryption.GetKeySchedule ()), + [in]"r"(in), [out]"r"(out) + : "%xmm0", "%xmm1", "memory" + ); } else #endif @@ -823,29 +835,29 @@ namespace crypto void CBCDecryption::Decrypt (int numBlocks, const ChipherBlock * in, ChipherBlock * out) { -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES if(i2p::cpu::aesni) { __asm__ - ( - "movups (%[iv]), %%xmm1 \n" - "1: \n" - "movups (%[in]), %%xmm0 \n" - "movaps %%xmm0, %%xmm2 \n" - DecryptAES256(sched) - "pxor %%xmm1, %%xmm0 \n" - "movups %%xmm0, (%[out]) \n" - "movaps %%xmm2, %%xmm1 \n" - "add $16, %[in] \n" - "add $16, %[out] \n" - "dec %[num] \n" - "jnz 1b \n" - "movups %%xmm1, (%[iv]) \n" - : - : [iv]"r"((uint8_t *)m_IV), [sched]"r"(m_ECBDecryption.GetKeySchedule ()), - [in]"r"(in), [out]"r"(out), [num]"r"(numBlocks) - : "%xmm0", "%xmm1", "%xmm2", "cc", "memory" - ); + ( + "movups (%[iv]), %%xmm1 \n" + "1: \n" + "movups (%[in]), %%xmm0 \n" + "movaps %%xmm0, %%xmm2 \n" + DecryptAES256(sched) + "pxor %%xmm1, %%xmm0 \n" + "movups %%xmm0, (%[out]) \n" + "movaps %%xmm2, %%xmm1 \n" + "add $16, %[in] \n" + "add $16, %[out] \n" + "dec %[num] \n" + "jnz 1b \n" + "movups %%xmm1, (%[iv]) \n" + : + : [iv]"r"((uint8_t *)m_IV), [sched]"r"(m_ECBDecryption.GetKeySchedule ()), + [in]"r"(in), [out]"r"(out), [num]"r"(numBlocks) + : "%xmm0", "%xmm1", "%xmm2", "cc", "memory" + ); } else #endif @@ -869,22 +881,22 @@ namespace crypto void CBCDecryption::Decrypt (const uint8_t * in, uint8_t * out) { -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES if(i2p::cpu::aesni) { __asm__ - ( - "movups (%[iv]), %%xmm1 \n" - "movups (%[in]), %%xmm0 \n" - "movups %%xmm0, (%[iv]) \n" - DecryptAES256(sched) - "pxor %%xmm1, %%xmm0 \n" - "movups %%xmm0, (%[out]) \n" - : - : [iv]"r"((uint8_t *)m_IV), [sched]"r"(m_ECBDecryption.GetKeySchedule ()), - [in]"r"(in), [out]"r"(out) - : "%xmm0", "%xmm1", "memory" - ); + ( + "movups (%[iv]), %%xmm1 \n" + "movups (%[in]), %%xmm0 \n" + "movups %%xmm0, (%[iv]) \n" + DecryptAES256(sched) + "pxor %%xmm1, %%xmm0 \n" + "movups %%xmm0, (%[out]) \n" + : + : [iv]"r"((uint8_t *)m_IV), [sched]"r"(m_ECBDecryption.GetKeySchedule ()), + [in]"r"(in), [out]"r"(out) + : "%xmm0", "%xmm1", "memory" + ); } else #endif @@ -893,34 +905,34 @@ namespace crypto void TunnelEncryption::Encrypt (const uint8_t * in, uint8_t * out) { -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES if(i2p::cpu::aesni) { __asm__ - ( - // encrypt IV - "movups (%[in]), %%xmm0 \n" - EncryptAES256(sched_iv) - "movaps %%xmm0, %%xmm1 \n" - // double IV encryption - EncryptAES256(sched_iv) - "movups %%xmm0, (%[out]) \n" - // encrypt data, IV is xmm1 - "1: \n" - "add $16, %[in] \n" - "add $16, %[out] \n" - "movups (%[in]), %%xmm0 \n" - "pxor %%xmm1, %%xmm0 \n" - EncryptAES256(sched_l) - "movaps %%xmm0, %%xmm1 \n" - "movups %%xmm0, (%[out]) \n" - "dec %[num] \n" - "jnz 1b \n" - : - : [sched_iv]"r"(m_IVEncryption.GetKeySchedule ()), [sched_l]"r"(m_LayerEncryption.ECB().GetKeySchedule ()), - [in]"r"(in), [out]"r"(out), [num]"r"(63) // 63 blocks = 1008 bytes - : "%xmm0", "%xmm1", "cc", "memory" - ); + ( + // encrypt IV + "movups (%[in]), %%xmm0 \n" + EncryptAES256(sched_iv) + "movaps %%xmm0, %%xmm1 \n" + // double IV encryption + EncryptAES256(sched_iv) + "movups %%xmm0, (%[out]) \n" + // encrypt data, IV is xmm1 + "1: \n" + "add $16, %[in] \n" + "add $16, %[out] \n" + "movups (%[in]), %%xmm0 \n" + "pxor %%xmm1, %%xmm0 \n" + EncryptAES256(sched_l) + "movaps %%xmm0, %%xmm1 \n" + "movups %%xmm0, (%[out]) \n" + "dec %[num] \n" + "jnz 1b \n" + : + : [sched_iv]"r"(m_IVEncryption.GetKeySchedule ()), [sched_l]"r"(m_LayerEncryption.ECB().GetKeySchedule ()), + [in]"r"(in), [out]"r"(out), [num]"r"(63) // 63 blocks = 1008 bytes + : "%xmm0", "%xmm1", "cc", "memory" + ); } else #endif @@ -934,35 +946,35 @@ namespace crypto void TunnelDecryption::Decrypt (const uint8_t * in, uint8_t * out) { -#if defined(__AES__) && (defined(__x86_64__) || defined(__i386__)) +#if SUPPORTS_AES if(i2p::cpu::aesni) { __asm__ - ( - // decrypt IV - "movups (%[in]), %%xmm0 \n" - DecryptAES256(sched_iv) - "movaps %%xmm0, %%xmm1 \n" - // double IV encryption - DecryptAES256(sched_iv) - "movups %%xmm0, (%[out]) \n" - // decrypt data, IV is xmm1 - "1: \n" - "add $16, %[in] \n" - "add $16, %[out] \n" - "movups (%[in]), %%xmm0 \n" - "movaps %%xmm0, %%xmm2 \n" - DecryptAES256(sched_l) - "pxor %%xmm1, %%xmm0 \n" - "movups %%xmm0, (%[out]) \n" - "movaps %%xmm2, %%xmm1 \n" - "dec %[num] \n" - "jnz 1b \n" - : - : [sched_iv]"r"(m_IVDecryption.GetKeySchedule ()), [sched_l]"r"(m_LayerDecryption.ECB().GetKeySchedule ()), - [in]"r"(in), [out]"r"(out), [num]"r"(63) // 63 blocks = 1008 bytes - : "%xmm0", "%xmm1", "%xmm2", "cc", "memory" - ); + ( + // decrypt IV + "movups (%[in]), %%xmm0 \n" + DecryptAES256(sched_iv) + "movaps %%xmm0, %%xmm1 \n" + // double IV encryption + DecryptAES256(sched_iv) + "movups %%xmm0, (%[out]) \n" + // decrypt data, IV is xmm1 + "1: \n" + "add $16, %[in] \n" + "add $16, %[out] \n" + "movups (%[in]), %%xmm0 \n" + "movaps %%xmm0, %%xmm2 \n" + DecryptAES256(sched_l) + "pxor %%xmm1, %%xmm0 \n" + "movups %%xmm0, (%[out]) \n" + "movaps %%xmm2, %%xmm1 \n" + "dec %[num] \n" + "jnz 1b \n" + : + : [sched_iv]"r"(m_IVDecryption.GetKeySchedule ()), [sched_l]"r"(m_LayerDecryption.ECB().GetKeySchedule ()), + [in]"r"(in), [out]"r"(out), [num]"r"(63) // 63 blocks = 1008 bytes + : "%xmm0", "%xmm1", "%xmm2", "cc", "memory" + ); } else #endif @@ -1285,9 +1297,9 @@ namespace crypto } }*/ - void InitCrypto (bool precomputation, bool aesni, bool avx, bool force) + void InitCrypto (bool precomputation, bool aesni, bool force) { - i2p::cpu::Detect (aesni, avx, force); + i2p::cpu::Detect (aesni, force); #if LEGACY_OPENSSL SSL_library_init (); #endif @@ -1297,7 +1309,7 @@ namespace crypto CRYPTO_set_locking_callback (OpensslLockingCallback);*/ if (precomputation) { -#if defined(__x86_64__) +#if (defined(_M_AMD64) || defined(__x86_64__)) g_ElggTable = new BIGNUM * [ELGAMAL_FULL_EXPONENT_NUM_BYTES][255]; PrecalculateElggTable (g_ElggTable, ELGAMAL_FULL_EXPONENT_NUM_BYTES); #else @@ -1312,7 +1324,7 @@ namespace crypto if (g_ElggTable) { DestroyElggTable (g_ElggTable, -#if defined(__x86_64__) +#if (defined(_M_AMD64) || defined(__x86_64__)) ELGAMAL_FULL_EXPONENT_NUM_BYTES #else ELGAMAL_SHORT_EXPONENT_NUM_BYTES diff --git a/libi2pd/Crypto.h b/libi2pd/Crypto.h index d4cbda97..1419293b 100644 --- a/libi2pd/Crypto.h +++ b/libi2pd/Crypto.h @@ -307,7 +307,7 @@ namespace crypto void InitNoiseIKState (NoiseSymmetricState& state, const uint8_t * pub); // Noise_IK (ratchets) // init and terminate - void InitCrypto (bool precomputation, bool aesni, bool avx, bool force); + void InitCrypto (bool precomputation, bool aesni, bool force); void TerminateCrypto (); } } diff --git a/libi2pd/Identity.cpp b/libi2pd/Identity.cpp index a4a9f716..3a659c11 100644 --- a/libi2pd/Identity.cpp +++ b/libi2pd/Identity.cpp @@ -803,29 +803,12 @@ namespace data XORMetric operator^(const IdentHash& key1, const IdentHash& key2) { XORMetric m; -#if (defined(__x86_64__) || defined(__i386__)) && defined(__AVX__) // not all X86 targets supports AVX (like old Pentium, see #1600) - if(i2p::cpu::avx) - { - __asm__ - ( - "vmovups %1, %%ymm0 \n" - "vmovups %2, %%ymm1 \n" - "vxorps %%ymm0, %%ymm1, %%ymm1 \n" - "vmovups %%ymm1, %0 \n" - : "=m"(*m.metric) - : "m"(*key1), "m"(*key2) - : "memory", "%xmm0", "%xmm1" // should be replaced by %ymm0/1 once supported by compiler - ); - } - else -#endif - { - const uint64_t * hash1 = key1.GetLL (), * hash2 = key2.GetLL (); - m.metric_ll[0] = hash1[0] ^ hash2[0]; - m.metric_ll[1] = hash1[1] ^ hash2[1]; - m.metric_ll[2] = hash1[2] ^ hash2[2]; - m.metric_ll[3] = hash1[3] ^ hash2[3]; - } + + const uint64_t * hash1 = key1.GetLL (), * hash2 = key2.GetLL (); + m.metric_ll[0] = hash1[0] ^ hash2[0]; + m.metric_ll[1] = hash1[1] ^ hash2[1]; + m.metric_ll[2] = hash1[2] ^ hash2[2]; + m.metric_ll[3] = hash1[3] ^ hash2[3]; return m; } diff --git a/libi2pd/api.cpp b/libi2pd/api.cpp index 3bfb801c..ad24519f 100644 --- a/libi2pd/api.cpp +++ b/libi2pd/api.cpp @@ -38,9 +38,8 @@ namespace api bool precomputation; i2p::config::GetOption("precomputation.elgamal", precomputation); bool aesni; i2p::config::GetOption("cpuext.aesni", aesni); - bool avx; i2p::config::GetOption("cpuext.avx", avx); bool forceCpuExt; i2p::config::GetOption("cpuext.force", forceCpuExt); - i2p::crypto::InitCrypto (precomputation, aesni, avx, forceCpuExt); + i2p::crypto::InitCrypto (precomputation, aesni, forceCpuExt); int netID; i2p::config::GetOption("netid", netID); i2p::context.SetNetID (netID); From c53ad2012ceb9c7489c97378e86500332d8edbac Mon Sep 17 00:00:00 2001 From: r4sas Date: Wed, 6 Sep 2023 20:26:46 +0000 Subject: [PATCH 5/7] [cpu] use cpuid on gcc < 5 Signed-off-by: r4sas --- libi2pd/CPU.cpp | 24 ++++++++++++++---------- 1 file changed, 14 insertions(+), 10 deletions(-) diff --git a/libi2pd/CPU.cpp b/libi2pd/CPU.cpp index d42f1ddd..14ce3788 100644 --- a/libi2pd/CPU.cpp +++ b/libi2pd/CPU.cpp @@ -9,12 +9,16 @@ #include "CPU.h" #include "Log.h" -#if defined(_MSC_VER) -#include - #ifndef bit_AES #define bit_AES (1 << 25) #endif + +#if (defined(__GNUC__) && __GNUC__ < 5) + #include +#endif + +#ifdef _MSC_VER + #include #endif namespace i2p @@ -26,18 +30,18 @@ namespace cpu inline bool cpu_support_aes() { #if (defined(_M_AMD64) || defined(__x86_64__)) || (defined(_M_IX86) || defined(__i386__)) -#if defined(_MSC_VER) - int cpu_info[4]; - __cpuid(cpu_info, 1); - return ((cpu_info[2] & bit_AES) != 0); +#if (defined(__GNUC__) && __GNUC__ > 4) + __builtin_cpu_init(); + return __builtin_cpu_supports("aes"); #elif defined(__clang__) #if __clang_major__ >= 6 __builtin_cpu_init(); #endif return __builtin_cpu_supports("aes"); -#elif defined(__GNUC__) - __builtin_cpu_init(); - return __builtin_cpu_supports("aes"); +#elif (defined(_MSC_VER) || (defined(__GNUC__) && __GNUC__ < 5)) + int cpu_info[4]; + __cpuid(cpu_info, 1); + return ((cpu_info[2] & bit_AES) != 0); #else return false; #endif From 4a5e16b99405182c6daa5862088a9b9247e1c6b4 Mon Sep 17 00:00:00 2001 From: r4sas Date: Thu, 7 Sep 2023 16:56:53 +0000 Subject: [PATCH 6/7] [cpu] tests with preprocessor Signed-off-by: r4sas --- libi2pd/CPU.cpp | 19 ++++++++++++------- 1 file changed, 12 insertions(+), 7 deletions(-) diff --git a/libi2pd/CPU.cpp b/libi2pd/CPU.cpp index 14ce3788..53771ff0 100644 --- a/libi2pd/CPU.cpp +++ b/libi2pd/CPU.cpp @@ -30,21 +30,26 @@ namespace cpu inline bool cpu_support_aes() { #if (defined(_M_AMD64) || defined(__x86_64__)) || (defined(_M_IX86) || defined(__i386__)) -#if (defined(__GNUC__) && __GNUC__ > 4) +# if (defined(__GNUC__) && __GNUC__ > 4) +# warning("CPU: IN GCC!!!") __builtin_cpu_init(); return __builtin_cpu_supports("aes"); -#elif defined(__clang__) -#if __clang_major__ >= 6 +# elif (defined(__clang__) && !defined(__GNUC__)) +# warning("CPU: IN CLANG!!!") +# warning(__clang__) +# if (__clang_major__ >= 6) __builtin_cpu_init(); -#endif +# endif return __builtin_cpu_supports("aes"); -#elif (defined(_MSC_VER) || (defined(__GNUC__) && __GNUC__ < 5)) +# elif (defined(_MSC_VER) || (defined(__GNUC__) && __GNUC__ < 5)) +# warning("CPU: IN MSVC!!!") int cpu_info[4]; __cpuid(cpu_info, 1); return ((cpu_info[2] & bit_AES) != 0); -#else +# else +# warning("CPU: FALSE") return false; -#endif +# endif #else return false; #endif From 4ed19c05f68e58a15c57f2791b1b0e1d863dda55 Mon Sep 17 00:00:00 2001 From: r4sas Date: Thu, 7 Sep 2023 21:20:52 +0000 Subject: [PATCH 7/7] [cpu] split AES detect for GNU C < 5 Signed-off-by: r4sas --- libi2pd/CPU.cpp | 31 ++++++++++++++++--------------- 1 file changed, 16 insertions(+), 15 deletions(-) diff --git a/libi2pd/CPU.cpp b/libi2pd/CPU.cpp index 53771ff0..106882eb 100644 --- a/libi2pd/CPU.cpp +++ b/libi2pd/CPU.cpp @@ -30,29 +30,30 @@ namespace cpu inline bool cpu_support_aes() { #if (defined(_M_AMD64) || defined(__x86_64__)) || (defined(_M_IX86) || defined(__i386__)) -# if (defined(__GNUC__) && __GNUC__ > 4) -# warning("CPU: IN GCC!!!") +#if defined(__clang__) +# if (__clang_major__ >= 6) __builtin_cpu_init(); +# endif return __builtin_cpu_supports("aes"); -# elif (defined(__clang__) && !defined(__GNUC__)) -# warning("CPU: IN CLANG!!!") -# warning(__clang__) -# if (__clang_major__ >= 6) +#elif (defined(__GNUC__) && __GNUC__ >= 5) __builtin_cpu_init(); -# endif return __builtin_cpu_supports("aes"); -# elif (defined(_MSC_VER) || (defined(__GNUC__) && __GNUC__ < 5)) -# warning("CPU: IN MSVC!!!") +#elif (defined(__GNUC__) && __GNUC__ < 5) + int cpu_info[4]; + bool flag = false; + __cpuid(0, cpu_info[0], cpu_info[1], cpu_info[2], cpu_info[3]); + if (cpu_info[0] >= 0x00000001) { + __cpuid(0x00000001, cpu_info[0], cpu_info[1], cpu_info[2], cpu_info[3]); + flag = ((cpu_info[2] & bit_AES) != 0); + } + return flag; +#elif defined(_MSC_VER) int cpu_info[4]; __cpuid(cpu_info, 1); return ((cpu_info[2] & bit_AES) != 0); -# else -# warning("CPU: FALSE") - return false; -# endif -#else - return false; #endif +#endif + return false; } void Detect(bool AesSwitch, bool force)